Structure and Dynamics of Diffusion Networks

Structure and Dynamics of Diffusion Networks

STRUCTURE AND DYNAMICS OF DIFFUSION NETWORKS A DISSERTATION SUBMITTED TO THE DEPARTMENT OF ELECTRICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY Manuel Gomez Rodriguez May 2013 © 2013 by Manuel Gomez Rodriguez. All Rights Reserved. Re-distributed by Stanford University under license with the author. This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/ This dissertation is online at: http://purl.stanford.edu/rq863nx7298 ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Andrew Ng, Primary Adviser I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Emmanuel Candes I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Jurij Leskovec I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy. Bernhard Scholkopf Approved for the Stanford University Committee on Graduate Studies. Patricia J. Gumport, Vice Provost Graduate Education This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives. iii Abstract Diffusion of information, ideas, behaviors and diseases are ubiquitous in nature and modern society. One of the main goals of this dissertation is to shed light on the hidden underlying structure of diffusion. To this aim, we developed flexible proba- bilistic models and inference algorithms that make minimal assumptions about the physical, biological or cognitive mechanisms responsible for diffusion. We avoid mo- deling the mechanisms underlying individual activations, and instead develop a data- driven approach which uses only the visible temporal traces diffusion generates. We first developed two algorithms, NetInf and Multitree, that infer the net- work structure or skeleton over which diffusion takes place. However, both algorithms assume networks to be static and diffusion to occur at equal rates across different edges. We then developed NetRate, an algorithm that allows for static and dy- namic networks with different rates across different edges. NetRate infers not only the network structure but also the rate of every edge. Finally, we develop a general theoretical framework of diffusion based on survival theory. Our models and algorithms provide computational lenses for understanding the structure and temporal dynamics that govern diffusion and may help towards fore- casting, influencing and retarding diffusion, broadly construed. As an application, we study information propagation in the online media space. We find that the informa- tion network of media sites and blogs tends to have a core-periphery structure with a small set of core media sites that diffuse information to the rest of the Web. These sites tend to have stable circles of influence with more general news media sites acting as connectors between them. Information pathways for general recurrent topics are more stable across time than for on-going news events. Clusters of news media sites iv and blogs often emerge and vanish in matter of days for on-going news events. Major social movements and events involving civil population, such as the Libyan's civil war or Syria's uprise, lead to an increased amount of information pathways among blogs as well as in the overall increase in the network centrality of blogs and social media sites. Additionally, we apply our probabilistic framework of diffusion to the influence maximization problem and develop the algorithm InfluMax. Experiments on syn- thetic and real diffusion networks show that our algorithm outperforms other state of the art algorithms by considering the temporal dynamics of diffusion. v Acknowledgement I would like to thank my advisor Andrew Ng for his advice and support. Hopping between Stanford University and MPI for Intelligent Systems would not have been possible without his help. I am deeply grateful to my co-advisor Bernhard Sch¨olkopf, who I consider truly a Doktorvater, for his mentorship, advice and support. I have learned a lot from his views on what is research all about, and the freedom and encouragement he has given me over the years have been exceptional. I would also like to thank my first reader Jure Leskovec for introducing me to research on networks. I truly admire his drive and passion. He has always managed to challenge me, and has been a great source of motivation and advice. It has been an honor to count with Emmanuel Candes as the second reader for this dissertation, and I would like to thank him for his advice and comments. I would also like to thank Christopher Potts for accepting the invitation to chair my orals committee. I would also like to thank my coauthors Andreas Krause, David Balduzzi, Jan Peters, Moritz Grosse-Wentrup, Jeremy Hill, Alireza Gharabaghi, Monica Rogati, Georgios Naros, and Jens Kober. I would like to thank Jan and Moritz for their men- toring and advice and colleagues from MPI for Intelligent Systems for great scientific discussions. I would like to thank my friends from Stanford (Borja, Gemma, Yiannis, Sotiria, Nicholas, Hylke, Ana, Amir, ...) for inspiring conversations, dinners, and trips. I would also like to thank my friends from T¨ubingen,specially my officemate Michel, for great endless conversations inside and outside the office, Tim, for his Californian vi optimism and his animal style American burgers, Shih-pi, for her support and uncon- ditional friendship, Ander, for keeping me a little bit fitter and always trying to keep me out of the office, and Lana, for her love, for helping me having a more balanced life, including becoming a tango dancer, and for being an adventurous traveler. I have great memories from my year in Limerick and would like to thank specially Karen, Alan, Gabi and Xavi, for making me having a great time despite of the weather. I am also very fortunate to have still many good friends in Pontevedra, my hometown, despite not spending much time there anymore: Yoli, both Miguis, Santi, Gon, Os- car, Robert, and many others, who always welcome me back as if time wouldn't have passed. Last but not least, I am very grateful to my parents Enrique and Margarita, for their unconditional love, kindness and values and to my sister Maria for always believing in me and my skills, even when I did not. Os dedico esta tesis a los tres! vii Contents Abstract iv Acknowledgement vi 1 Introduction 1 1.1 Inference of diffusion networks . .2 1.2 Influence maximization in diffusion networks . .4 1.3 Thesis organization . .5 2 Background and basic concepts 8 2.1 General network-theoretic concepts . .8 2.2 Models of network structure . 10 2.2.1 Forest Fire Model . 11 2.2.2 Kronecker Graph model . 12 2.3 General diffusion-theoretic concepts . 14 2.4 Sets, functions and distributions . 17 2.5 Real data . 19 2.5.1 Memes and hyperlinks . 19 2.5.2 MemeTracker dataset (2008/09/01 ! 2009/08/31) . 20 2.5.3 Topic-based dataset (2011/03/01 ! 2012/02/28) . 21 2.6 Table of symbols . 23 3 Survey of related work 24 3.1 Inference of diffusion networks . 24 viii 3.2 Influence maximization in diffusion networks . 26 4 Inference of unweighted diffusion networks 28 4.1 Introduction . 28 4.2 Problem formulation . 30 4.2.1 Data . 31 4.2.2 Pairwise interactions . 32 4.2.3 Likelihood of a cascade for a given tree . 33 4.2.4 Probability of a tree in a given network . 34 4.2.5 Likelihood of a cascade in a given network . 34 4.2.6 The unweighted network inference problem . 35 4.3 NetInf .................................. 36 4.3.1 Algorithm . 36 4.3.2 Experiments on synthetic data . 49 4.3.3 Experiments on real data . 61 4.4 Multitree ................................ 68 4.4.1 Algorithm . 68 4.4.2 Experiments on synthetic data . 72 4.4.3 Experiments on real data . 75 4.5 Summary . 76 5 Inference of weighted diffusion networks 78 5.1 Introduction . 78 5.2 Problem formulation . 80 5.2.1 Data . 81 5.2.2 Pairwise interactions . 82 5.2.3 Probability of survival given a cascade . 83 5.2.4 Likelihood of a cascade . 84 5.2.5 Relation to unweighted networks . 85 5.2.6 Three weighted network inference problems . 86 5.3 NetRate ................................. 88 5.3.1 Algorithm . 88 ix 5.3.2 Properties of NetRate ..................... 90 5.3.3 Solving NetRate ........................ 91 5.3.4 Experiments on synthetic data . 97 5.3.5 Experiments on real data . 113 5.4 Summary . 123 6 Influence maximization in diffusion networks 126 6.1 Introduction . 126 6.2 Problem formulation . 128 6.2.1 Pairwise interactions . 128 6.2.2 Continuous time diffusion process . 129 6.2.3 Continuous time influence maximization problem . 130 6.3 InfluMax ................................ 131 6.3.1 Influence evaluation . 131 6.3.2 Influence maximization . 134 6.3.3 Experiments on synthetic data . 137 6.3.4 Experiments on real data . 145 6.4 Summary . 146 7 Diffusion and survival theory 149 7.1 Introduction . 149 7.2 Diffusion as a counting process . 150 7.3 Additive risk model of diffusion .

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    195 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us